REVISTA MATEMATICA de la

Uiversidad Complutense de Madrid

Volumen 8, nmero 1: 1995

The Initial Value Problem br the Equations

of Magnetohydrodynamic Type in
Non-Cylindrical Domains
M.A. RQJAS-MEDAR aid R. BELTRAN-BARRIQS

ABSTRACT. By using the apectral Galerkin method, we prove the existence of weak solutions for a system of equatics of magnetohydrodynamic
type ji non-cylidrical domains.

1. INTRODUCTION
In several situations the mation of incompressible electrical conducting flnids can be modelled by the so called equations of magnetohydrodynamics, which correspond to the Navier-Stokes equatians coupled
with the Maxwells equations. Iii tlie case where there is free motion of
1991 Mathe,natics Subject Classiflcation: SSQiO, 76D99, 35005
Servicio publicaciones Univ. Complutense. Madrid, 1995.

230

MA. Rojas-Medar and R. Beltrn-Barrios

heavy ions, not directly die to the electric fleid (see Schliiter [14] and
Pikelner [12]), these equations can be reduced to the following form:

%~

tu + u.Vu

pa~~+
1

uXlh

$It.S7It
h.Vu =

9V(p + ~
grad w,
=

div u = 0,
div h = 0,
together with suitable boundary aid initial conditions.
Here, u and It are respectively the unknown velocity and magnetic
fields; p is the unknown hydrostatic pressure; w is an unknown function related to the motion of heavy jons (in such way that the density
of electric current, Jo, generated by this motion satisfles the relation
rot Jo = aS7w); Pm is tite density of mass of tite fluid (assmed to
be a positive constant); ~ > O is tite constant magnetic permeability of
tite meclium; u > O is tite constant electric conductivity; i~ > 0 is tite
constant viscosity of tite fluid; f is an given externad force fleld.
In titis paper we will consider tite problem of existence of weak
solutions for tite problem (1.1) in a time-dependent domain of IR x
(0, T),n=2,0< T<+oo.
To (1.1) we append the following initial and boundary conditions

uLaq
u(O)

and

haQ = O

for ah t,

(1.1)

uo

aid

It(O) = It0

in Q(0),

(1.2)

where uo and Ito are given functions. In (1.1), tite diferential operator
V aid t are tite usual gradient aid Laplace operator, respectively.
Tite main goal in this paper is to show existence of weak solutions
for the initial value problem (1.1)-(1.3). Qur strategy for setting titis
question consists of transforming problem (1.1)-(1.3) into anotiter initial
boundary-value problem in a cylindrical domain whose sections are not
time-dependent. This is done by means of a suitable change of variable.

The 1iti,l value Problem for the Equatio....

231

Next, this new initial value prablem is treated using Galerkin approximation. We conclude returnung to
using the inverse of the aboye
change of variable. This technicaiity was introduced by Dal Passo aid
Ughi [4] to study a certain class of parabolic equations in non-cylindricai
domains.

We feel that it is apprapriate to cite some earlier works on the initial

value problem (1.1)-(1.3) aid to locate or contribution therein. The
cylindrica.l case of (1.1)-(1.3) has been studied by sorne authors. Among
them, let us mention the paper of Lassner [7], Boldrini aid Rojas-Medar
[2], Rojas-Medar and Boldrini [13].
Lassner [7] by making the use of semigroup techniques a.s the ones in
Fujita and Kato [5] to show the local existence and uniqueness of strong
solution. The more constructive spectral Galerkin method was used by
Boldrini and Rojas-Medar [2], [13] to obtain local, global existence and
uniqueness of strong solutions. The aboye authors working in IR with
it = 2 or 3.
A mathematical study of the problem (1.1)-(1.3) in anon-cylindrical
domain was not done, however, it has to be pointed oit that similar timedependent problem bit for tIte Navier-Stokes and Boussinesq problems
have been studied by many different authors. This is the case, for instance of tIte works Lions [9], [8], Fujita and Sauer [6], deda [10], Qtani
and Yamada [11],
Conca and Rojas-Medar [3]. In particular, we would
like to emphasize that the arguments of tIte mentioned authors demand
that Q(t) be nondecreasing with respect to t (see Lions [9],
problem 11.9,
p. 426); except tIte work of Conca aid Rojas-Medar [3].
In this work, we wili adapt the techniqe by [3]to tIte system (1.1)-

(1.3).
TIte paper is organized as foliows. Iii Section 2 we introduce varjois
functional spaces and state the theorems. Section 3 aid 4 deal with their
proofs.
2. FUNCTION SPACES AND PRELIMINAIRES
TIte functions in tIte paper are elther IR or IR-vauued aid we will
not distinguish these twa situations in or notations. To which case we
refer wifl be clear from the context.

232

MA. Rojas-Medar sud R. Beltr-B,rrios

Let =2be a baunded domain in IR (n =2) with smooth boundary

0=2.We write simply hill for L2(=2)-norm.TIte inner product in L2(=2)
is denoted by (u, y). The solenoidal function space is defined as usual
C&2=2)= {~ 6 Cr(=2) div cp

fi(=2)= the completion of C3~

2(=2) norm,

0}=2)under the L
V
3(=2)= the completion of C0%(=2)under the fiS(=2) norm

where H(=2)denote tIte usual Sobolevspace with s 6 la.

The norm and mier praduct in fi(=2)and V~(=2)are:

(1~9)

(u, y).

and

f19dx,

fil

V)H

IIuu~

(u,u)V

=1

where (u~, v,)H is tIte standard inner product of fi5(=2);(V,(=2))denote6

tIte toplogical dual of V,(=2).
bu particular, we denote
V
1(=2)= V(=2) aid

mili

Vu.

We will use other standard notations aid terrninology; for tliem,

we refer to Adams[1] aid Temam [15].
Let r be a real-valued fuiction which is of C-class on tIte interval
[0,2] such that,
rQo)

min{r(t) O =1 =2}

(this candition is essential).

>

(2.1)

233

The Initial Value Probletn for the Equatios...

QQ)

TIte time-dependent space domain

is a bounded set
defined by
Q(t) = {x E IR ~1 < r(l), 0= 1 =T}
where

iii

IR

denotes tIte usual norm u IR. Its boundary is

OQ(t) = {x E IR xI

= i(t),

0= 1 =T}.

It is warth noting that such domains Q(t) O < t

non-cylindrical time-dependent domain of IR x IR:

<

2, generate a

Q= O<t<T
U Q(t)x{l}.
In sucIt conditions, we can now define a notion of weak solution for
(11)-(13):
Definition 2.1.
defined

We sItal say 1Mb a couple of funclions (u,It)

un Q

is a weak sohution of (1.1)-(1.8,) 1ff

(1) u,h E L2(O,2;V(Q(t))) fl L00(O,2;H(QQ)))
n

It

(Ii) f~ [au.sotu >j Vu 7p+a

1S
af~fsodxdt,

>~

(iii) f~ [h.@ty E VhV~

1+ ~

=1

i;=1

Br;

u SItj

h.]dxdt =

~
Ox;

u~ ~f It1

2Iti~f

u]dzdi = O

foi ah ~
E C(Q) wilh div ~ = div ~ = O, tIte suifiz t denotes bIte
operaior ~ (deritatives witIt respeet lo 1 ull sometimes also be denoled
by a or simply by d/dt).
(iv) u(O) = no,
mhere pu a =

It(O) = It0.
u = p and y
~.

= pu

Remark 2.2. Ii this definition tIte initial conditions (iv) Itave the
usual meaning; see for exaznple, Lions [8].
TIte main result of our article is

234

MA. Rojas-Medar sud R. Beltrn-Barrios

Theorem 2.3. 1ff E L2(Q), u0, It0 E H(Q(O)), bIten lhere crisIs
a u,eak solulion of problera (1.1)-(1.8) for any lime interval [0,2].
Theorem 2.4. If it 2 tIte solubion (u, h) obtained un lIte TIteorera
2.8 is unique. Moreoter u and It are almost everywItere equal bofunetions
eontinuous frora [0,2] mb fi and

u(t).uo

irz.H,

ast*O

(2.2)

It(t)*Ito

infi,

ast*O.

(2.3)

3. PROOF OF THEOREM 2.3.

Let us introduce the traisformation

it

U, givenby

1where U which
D x verifies
(0,2) aid
D =we{xseeE easily
IR. 1 x
=1}.
r(t) isa C
fuiction
(2.1),
that
y is Since
a diffeomorpItism
aid that its inverse r1 : U
Q satisfles

r(y, t) = (r(t)y, t).

(3.1)

We also define
v(y,i)

b(y,i) = h(yr(t),t),
q(y,t) =

E(y,t) = w(yr(t),t),
J(y,t) = f(yr(t),t),
vo(y) = uo(yr(O)),
bo(y)=ho(yr(O)).

(3.2)

235

The Initial Value Problem <oc the Equations...

By using (3.2), tIte system on

Q (1.1)-(1.3) is transformed lito tIte

system:

tv+

Z
r(t)

9v

Ob

oye

.~

(3.3)
r(t) ~ 8v

/1

0~ + r(t)

0v

Ob

r(t)

4t)Z
(3.4)

t,o(y)

(3.5)

b(O,y) = bo(y)

(3.6)

v(0,y)

on tIte cylindrical domain U = D

><

(0,2).

Qn tIte other hand, let us set

c(v,w)=

Dv1

~j=i

a(v~w)=J

B(u,v,w)

9vj

Otiij

oyi
0v1
w1dy

for vector-valued functions u, y aid w for whicIt tIte integrals are well-

detined.

236

MA. Rojea-Medar ad R. Beltr-Barrios

The notation of wealc solution for (3.3)-(3.6) is completely similar

to tIte ones for (1.1)-(1.3).
To prove tIte existence of solutions of tIte transformed system (3.3)(3.6) we will use tIte spectral Galerkin method. That is, we lix .s
and we consider the Hilbert special basis {w(xfl421 of T~8(D), whose
elements we wilI choose as the solution of tIte following apectral problem
(w~,v)~ =

A(u9,v)

Vv 6 1/8(D).

Let Vk be the subspace of 18(D) spanned by {w,. .. ,wk}. For

every k =Y, we define approximations ~ b of ti and b, respectively, by
meais of the following hite expansions:
k

yk

>j3 C.k(t)W(X)

6 Vk

t E (0,2)

Zdk(t)u,t(x) 6 Vk

te (0,2)

i=1

aid
k
=

i=1

where the coefficients (elk) aid (d~k) will be calculated in such a way
that y and b solve tIte following approximations of system (3.3)-(3.6):
1/

a(vk

r(t)

~pjt

r(t)
r(t)

(3.7)

(b4)+ [r(t)]

a(b,iP)+

r(t)

r(t) c(b

1
r(t)

B(v,b,t,b)

1
rQ)

B(bk,vk,t)

(3.8)
~

237

The Iitial V,lue Problem for the Equstio....

for adJ ~,4, 6V,

y

where vt

. y

aid b~

(O) = t4; b(O)

b0

ji

(3.9)

H(D) as k -+00.

Equatians (3.7), (3.8) aid (3.9) is a system of ordinary differetial

equations for tIte coefficient functios ei~(t) aid d1~(l), whjch defines
0< aid bk in a interval [O,1,4. We will show some a priori estimates
independent of k aid 1, u order to take 4 = 2. Also, we will prove that
(yk,bk)
converges u appropriate seise to a sohution (v,b) of (3.3)-(3.6).
We prave tIte followig lemma.
Lemnia 3.1. TIte transformed
system (8.8,)-<8.6) admts ab leas
2(0,2;V(D))flL00(O,T;fi(D)).
one weak sohulion (v,b) un L
Proof. Settig ~ = 0< aid 4,
bk u (3.7) and (3.8), respectively,
we Itave

ad
2 di

y 112+

Il~vIl2

l/

[r(t)]2

a(J, vk) +
d

~ bj2 +

[r(t)]2

B(bk, bk, yk) +

r(t)

vM2

1
=

r(t)

B(b

ar(t
r(t)

e(vk, y)

~Q)

yk,bk) + ~

r(t)

le

c(b b)

since B(v, tu, tu) = O for tu E y

Adding tIte aboye inequalities aid observing that
r&) [B(b,b,v) + B(b,v< bkfl

(allvklI2 + lbkll2) +

cr(J,v )+ r(t)
=

EF.

O, we obtain

[rQ)]2

(vIIVv112 + yVb2)

[ae(v ,v)+c(b,b)]

238

M.A. Rojas-Medar sud R. Beltrn-Barrios

Now, we use Hlder aid Young inequalities to estimate tIte rightItaid side of tIte aboye inequality, we obtain

lFkl =~ 1h112 + ~ IhvlI~ + C~r(t)h2a2Iy1~ 0<112

+ Cr(t)2yIj.ob2 +

Vv 2

[(t)]2

+ [(t)]2
wItece, taking

/2 aid 6

~ (a~v~ + jb2) +

=]~i2 +
<

CIJ12

-y/2, we arrive at tIte inequality

(uVv2

(cr;(02a211Yioo +

+ yVb2)

110<112 + cr(t)fly~.~ hlbIr

+ C(alv112 + lIb112),

where C is a positive constait that depends oly max {y(t) 0 < 1 _

a,
By integratig the aboye iequaiity between O aid 1 with O
we conclde:
+ jb(t)12 +

[r(s)]~

=cft jJ(s)2ds + C

(vVv(s)2

(~yk(s)2

allvk(O)112 +

<

1 <2

+ yllVb(s)ds

b(s)jj2)ds+

The Initial Value Problem for the Equ,tios...

that

239

Dueto the choice of 4 and b~, there exists C independent of k such

1411 =Cvo, [b~ =Cjbo, moreover f IIJ(s)112d8 is limite for

O < t < 2, we conclude, by usig Growalls inequality, that

2 +
+ b(t)

(vVv(s)2 +

yIhVb(s)112)ds =e

1.

TItus, for al k we Itave that 0 and b exists globally

in 1 aid
2(O,2;V(D))
as
aid TIte
(b)next
remains
bouded
in consists
L00(O,2;fi(D))
rl that
L (v~) aid (b~)
00.
step of
tIte proof
of proving
bouded in L2(O,2; (1
are
8(D))). lo tItis end, let us flx some otations
(vk)

<A(t)v, u>

2 a(v, u)
[rQ)]
rQ) e(v,u)

<C(t)v, u> =

B()
<E(flv, u> =
(C(tft#9, (E(t)vk) and (E(flb) are
We wili prove
that (Q)v),
2(0,2;(V
bounded in L
5(D))).
Indeed, for ah u 6 14, we have
1
IA<A(t)v,u>I = [r(t)]2

[r(t)]2

whence, we have

IIA(t)vII(v(D))v =

sup

I<A(t)vk,u>l

240

M.A. Roj,s-Medar ad R. Beltr-Barrios

cansequently

6 C([O,2]) and (y) is bounded in L2(O,T;V(D)), we

Since
deduce that

[r(t)]2 Slv(t)j2dt < L

and (A(t)v) is therefore bouded jn L2(O,2;(V.(D))). Aadogously,
we shaw that (C(t)v) js bounded in L2(O,T;V
8(D))). weTowill
prove
2(O,2;(V~(D)))
use the
tIte
boundedness
of
(E(t)v)
jn
tIte
space
L
following interpolation result whose proof can be faunA lxi Liana [8]:
Lemma 3.2. If (y) is bounded in
L2(O,2;V(D)) flLw(O,2;fi(D)),
tIten (y) is also bounded un L4 (O,2;LP(D)), u,Itere 1
~>

= 12

1
2n

Using thjs Lemma, we conclude that

r(t)

_ r(t)

h114IILPIIViIILPIIyII
y

since 1 + 1 + = 1 and tIte Soholev embedding

1
p
TItis imply

fi91(fl)

JI

E(t)v~, (D>),dt <

fT

[r(t)]2 v(b)lt~dt

L(D).

The Initial Value Problem ter the Equatio....

241

Sjnce
E C([O,2]), we ca conclude tItat (E(t)v) is boinded
2(0,2;(14(D))).
ji L
Similary, we prove tItat (E(t)b) is bounded u
Now, we consider P,: fi

y defined by
le

P~u

Z(u,wt)w
t=1

14(D) c fi aid y C 14(D), we can consider F,, : 14(D)

14(D). Itis easy to see tItat Pi, E L(14(D),V
8(D)), (L(X,Y) denote
the space of tIte bounded operator of X into Y) hence
*

lies in L((V8(D)), (14(D))) and

defined by <PZ(v),w> =
lPk~II =FkI =1.
We also observe tItat tIte autofunctions w are invarjants by Pi,, i.e.,
1.
Pi,(wt)

From jt and (3.7) we conclude that

a(vP,w)

<(VA(t)

<PZ((vA(t)

cxE(t)

aC(t))v + QJ +

aE(t)

aGY(t))0< + aJ +

aC(i))v + aJ + E(t)b),w>.

whence, for ad w 6 V~, we Itave

cx(vflw) = <PZ((vA(t)

aE(t)

Hence, by taking w = Pi,u, for u 6 14(D), we obtain

a(vflu)

<P,((VA(t)

aE(t)

aC(t))v + cd + 12 (t)b),u>,

242

MA. Rojas-Medar ad R. Beltr-Barrios

and, consequently
e
= P,((vAQ)
aE(i) cve(t))v + aJ + E(t)b)
2(O,2;(14(D)))
thaks
to the previous estimates aid
bebong to L
Ff1 =1.

av~

Working as before, we Itave

=

PZ((yA(t)

e(t))b + fi(v,b)

fi(b,ve)),

where fi(u,w) = ~r~) (u,w) aid <(u,w),h> = B(u,w,h). Consequently, it js sufficient to show tItat H(v,b) and fi(be,v) belong to
L2(O,2;(14(D))) to conclude tItat b~ is boinded u L2(0,2;(14(D))).
We observe that

JI

IlH(v,b)II~v(D>ydt

Lr(t)

Qn tIte other hand,

I<(ve,b),It>l =>3
1,1

v~ IL~ b~ LP II ~I Ln
01/1

=C 1v LP lIbIILP jIt

8,

aid therefore
(v, b)Il~V(D)) < Cv~ jbj,

243

The Initial value Prblem for the Equatione...

tItis imply
T

c(fT

[J)]2ve~y)I/2(fT

[(i)]2 b

Similarly, we prove that fi(b,v) is bounded u

llt~)

<

L2(O,T;(V,(D))).

Therefore, arguig as ji tIte book of Lions [8, p. 76] and

making use of the Aubin-Lios Lemma with B
0 = V(fl), Po = 2, -B~ =
(14(D)), Pi = 2 aid B = fi(D) (see Theorem 1.5.1 aid Lemma 1.5.2
of2(O,2;V(D))
tIte aboye baak,
conclude that
exist denote
y, b 6
such p.
tItat,58),
up we
to acan
subsequence
wItichtItere
we sha]I
L
again by the suffix le, tItere hold
U

le

*271

b ...4bj

: }
le

~ *v~
bt
-+

in L2(O, 2; V(D)) aid L00(O,2;H(D)) wealdy aid

L2(O,T;fi(D)) strongly, as k

in L2(O,T;(V

8(D)) wealdy, as le 400.

Now, tIte ext step is to take the limit. But, once tIte aboye con
vergence results Itave been establisIted, tItis is standard procedure aid

it follows the sanie patter as u

Lions [8, p. 76-77]. Consequetly, we

244

MA. Rojas-Medar ami R. Beltrn-Barrios

shail omit jt and we wili directly deduce that

2

-f
JIo 1.

a(v,ek) +

r(t)

[r(i)]2 a(v, #) +

B(b b, ~)

fT

(3.10)

a(v,4,) +

r(t)

~T

a(J, ~) +

aid

fi

5
-y
a(b, 4,) +
[r(t)]2

c(v,t)

lo r(t)

(3.11)

r(t)
c(b,~)
o r(t)

JI
for ah 4>,4, 6 <21(U) such tItat div ~
proved.

so

div 4,

O. So, the Lemmais

To conclude the proof of TIteorem, let us consider a test function

E C1(U) such tItat div s~ = O, and define
y,t)

[r(l)]~so(yr(t),t).

Iiitegrating by parts,

a(v,qS)

f2 aEjj]ev~o=f

~[r(j)]fl(y, so)

245

The Initial Value Problem for the Equation....

and also
i

JIfa

Os

Oyj 01/1

r(t)

>3

B(b,b,#) =

>3

Li JIi~~~~1

Jor(t)

0S~
O1/i 27~

ID a[r(t)fv,

11
]or(t)

01/1

By using the aboye jdentjtjes jn (3.11), we obtain,

JI

a[r(t))2v(y, t)p(yr(t), t) +

Ok

TI

->3
t=1
= T

>3

iJI

[r(t)]
r(t)

0v 0so
01/1 01/1

J~j

L IDarit~nvivi
JI a[r(t)]~Jso.

(3.12)
Analagously, we obtain for b,

fi

JI

~9JID
i

where

1 JI~i

O@j+

~ 6 ((U) with div ~

fi JIi~~

-y [r(t)]~ Ob 0~
r(t) Oy~ Oy~

0@

=0
81/1
(3.13)

O.

246

MA. Rojas-Medar ad R. Beltr-Bnrios

Q which js
Let us now consider tIte traisformation i-1 : U
defined by (3.1). We observe that its Jacobian is [r(t)]TI. Consequently,
by change of variables ji tIte integrals, (3.13) uand (3.14) become
*

1 a ~so
+ ~

JI

u1

vVuJLp~

>3.12

Ox5

1,1=1

+tJI

It5

JI

and

JI

It.

+f j~ yVIt~S7~1

>3

1,5=1

JI

TI

ni

Ox5

h.+>3

JI

It~

0@i
Ox5

~1 =

of the problem; since tIte

wItich proves that (u, It) is a weak solution
mappigs
2(0, 2; V(D))
L2(O, T; V(Q(tfl)
L
*

b(y,t)

-.4

u(x,t)

>

h(x,i)

___

aid

L2(0,2;H(D))

L2(0, 2; H(Q(t))

v(y,t)

n(x,t)

b(y,t)

>

It(x,t)

r(t)1)

b(),t)

247

Tbe lniti,I Value Problem for the Equationa...

are smooth bijectios of class C1, it follows that

u,It E L2(O,2;V(Q(t))) nL00(O,2;H(Q(t))).

Finally, a standard arguments sItows that u(O)

u
0 aid It(O) = It0

(see remark 2.2). TItis finished tIte proof of tIte Theorem.

4. PROOF OF THEOREM 2.4.

We first prove tIte regiilarity result.

Wey);
observe
that tIte applying
proof of
2(0, 2;
consequently,
tIte
aboye
theorem
sItown
that
u
E
L
Lemma 1.2 jn Teman [15], p. 260, we obtain that u is almost everywhere
equal to a fnction continuous from [0,2] lito fi.
Thus,
u E C([O,2];H)
and (2.2) follows easjly. Analogously it is proved tIte continuity of It aid

(2.3).
We also recail that Lemma 1.2 u Temam [15], p. 260-261, asserts
that tIte equatios below holds:
d
~

uQ)2

2<u(t), u(t)>,

TItese results will be used ji tIte followig proof of uniqueness wItich we

will start ow.
Consider that (ui,Iti) and (u
(1.1)-(1.3) with tIte same

2, It2) are two solutjons of tIte problem

aid uo,Ito aid define tIte differences w =

248
ni

MA. Rojas-Medar ad II. Beltrn-Barrios

u2 aid y

It1

It2. They satisfy

a(we,~) + va(w,#)

aB(w,u1,~b)

aB(n2,w,#)

+ B(y,It,~) + B(It2,v,~)
(vt,4,) + ya(v,4,)

B(u1,v,4,)

B(w,It2,4,)

+ B(v,n1,4,) + B(It2,w, 4,)

for any ~,4,CV; also w(O)

y(O)

0.

2(0,T;V);
By
the
proof
of
Theorem
2.3,
tut
aid
vj
belong
to
L
consequently by setting ~ = tu and 4, = y ji the aboye inequahities, we
obtain
ad

w +va(w,w)

~-

aB(w,u,w)+B(v,Iti,w)

+ B(It
2,v,w)

~ VII + -ya(v,v)

B(w,It2,v) + B(v,ui,y) + B(It2,w,v)

thanks to the aboye remark.

Addig the aboye identities, we get

d (ahw112 +

vlhwh + yjlv
(3.14

aB(w,ui,w) + B(v,It,w)

since B(It2,v,w) + B(It2,w,v)

O.

B(w,h2,v)+ B(v,ui,v)

249

me liti,I Value Problen, for Ihe Equatio....

I4ow, we observe that

aB(w,u1,w) =aCwjj5~huui
< a~Cjwj

Ihwlli Ilui Iii

~ IIwIh + C,.4a)aIIwII2IIui II~

where we used the Lemma 3.3 u
Hlder aid Young inequahities.

Temam [15], p. 291, together with

Aalogously, we can prove

B(v,ui,v) =~ t4~2 + C.,Iyl2Iluih~,

B(v,Iti,w) =~ lwl~ +

2)ItItill~

Iy[~ + C,c,(a)(Ilv12+ aIlwI

B(w,h
2,v) =~ w~ +

~ hhvll2 + C~,~(a)(lvlI2 + awj2)h

By using tite aboye inequalities u

2~.

(4.1), we get

w (a1w112 + IIvI2) + vuIwIi~ +

rUvII~
< C~(aw2 + IIvhl2)(hhuil~ + h1j~ + h2j~),

where C is a positive costant that oly depend on u, -y, a.

By integratung u
2
aw(fl

time, the use of Gronwahls iequality, we obtadn

+ hIvclt)l12 =(aw(0)l2 + lIv(O)ll2)e~(t>

250
whereso(t)

MA. Rojas-Medar and R. Beltrn-Barrios

=

(f~(ui~+It~+It2flds

+oo,foreveryt 6 [0,2].
This last iequality, implies w(t) = v(t) = O. Hence u1 = u2 and It1 = It2
aid the uniqueness is proved. This completes the proof of the Theorem.
<

References

[1] Adams, R.A.: Sobolev Spaces, Academic Press, New York, 1975.
[2] Boldrinj, J.L. and Rojas-Medar, M.A.: Qn a system of evolubion
equations of magnebohydrodynamic bype. To appear ji Matemtica Contemporanea.
[3] Conca, C. and Rojas-Medar, M.A.: TIte inilial value probhem for
lIte Boussinesq equabions in a lime-dependen domain, Informe Interno,
M.A.-993-B-402, Universidad de Chile, Chile, 1993, submitted.
[4] Val Passo, R. et Ughl, M.: Problme de Diriehel pour une classe
dquations parabohiques non hinaires dans des ouverts non cyhindriques,

C.R. Acad. Sc. Paris Srie 1, 308 (1989), 555-558.

[5] Fujita, H. aid Kato, T.: Qn tIte Navier-Stokes

muid

value problem,

1, Arch Rational Mech. Anal., 16 (1964), 269-315.

[6] Fujita, II. and Sauer, N.: (onstruelion of weak solutions of lIte
Navier-Sbokes equabions in a non-eylindrical domain, hill. Amer. Math.
Soe., 75 (1969), 465-468.
[7] Lasser, O.: Uber ein rand-anfangswerl-probhem der magnebohydrodinamik, Arch. Rationad Mech. Anal., 25 (1967), 388-405.
[8] Lions, 3.L.: Une remarque sur les problmes devohubion non hinaires
dais des domains
eglindriques, Rey. Roimaine Math. Pires Appl.,

non

9 (1964), 11-18.
[9] Lios, J.L.: Quelques mihodes de rsohution des problmes auz hmiles non-hineaires, finad Gaithier-Villars, Paris, 1969.
[10] eda, K.: Qn bIte inibial problem br lIte heal conveebion equation
of Boussinesq approximabion in a bime-dependent domain, Proc. Japan
Acad. Ser. A. Math Sci., 25 (1988), 143-146.

u noncyhindrical domains: Arz approach by tIte subdifferenlial operabor bIteory,

[11] tanj, M. aid Yamada, Y.: Qn lIte Navier-Sbokes equations

Fac. Sci. Unjv. Tokyo Sect lA, 25 (1978), 185-204.

The Initial Value Problern for the Eqnations...

251

[12] Pikelner, S.B.: Gnundiander den Kosmisc/zen dllcbrodynamic, Moscou, 1966.

[13] Rojas-Medar, M.A. aid Boldrini, J.L.: Global stmng sohutions of
equations of magneboItydnodynamic lype, Relatrio de Pesquisa,

R.P. 53,

IMECC-UNICAMP, Brazil, 1993, submitted.

[14] Schliiter, A.: Ognamie des plasmas, 1 aid II, Z. Naturforsch. Sa,
(1950), 72-78; 6a., (1951), 73-79.
[15] Temam, 11..: Navier-Sbokes equations, North-Holland, Amsterdam,
Rey. Edit., 1979.

UNICAMP rMECC, CPSO6S

Departamento de Matemtica,

13081-970, Campinas, SP

Universidad de Tarapach

Brazil

Casilla 7D, Anca

Chile
Recibido: 27 dc Octubre de 1993